Abstract

This thesis presents a numerical and experimental investigation of predictive control with different building-integrated thermal storage systems. The purpose is to utilize efficiently the available thermal mass of a space in order to reduce the peak demand and energy consumption while maintaining occupant comfort.
Two case studies are performed together with model development. The first one is a zone of a perimeter office with floor heating, convective cooling and a carpeted concrete floor. Various floor configurations are simulated to study the effect of thermal mass and its cover in predictive control. The window to wall (WWR) ratio is 75%, resulting in high solar gains. These solar gains are useful in winter but in summer they need to be reduced by controlling the shades to reduce total cooling load. The simulations results show that an exposed concrete floor (no carpet) can lead to significantly lower peak (about 35% in cooling and 41% in heating) in the auxiliary cooling/ heating as compared to a carpeted floor.
The second case study presents a simulation and experimental implementation of simple predictive control strategies in a test cell using a phase change material (PCM) as a means of thermal storage exposed to simulated outdoor conditions in a large environmental chamber. The PCM, embedded in a wall of the test cell, is actively charged through forced air circulation. The objective of the study is to investigate how model predictive control (MPC) can be used to leverage the thermal capacity of a PCM wall. This study also shows how a low-order thermal network model can be used as an effective tool in the design of the MPC strategy. The proposed MPC algorithm uses a set of linear ramp functions to change the room temperature setpoint to reduce and shift peak power demand. These ramp setpoint profiles allow the effective charging and discharging of the wall-integrated PCM. The algorithm applied in the experimental facility uses the outdoor temperature as input to select the best charging and discharging rates over a prediction horizon. A low-order model of the room and the PCM wall is used in the predictive control algorithm. It was found that this model can predict accurately the peak power demand and the temperature profile in the room. As the process moves forward in time, the weather profile is updated periodically and the algorithm calculates the new outputs over the new control horizon. The whole procedure is automated and the outputs of the algorithm are transferred to the local variables of the controller of the test room through BACnet.